1. INTRODUCTION

Previous analysis on german (left) and french (right) newspapers has demonstrated the interest to analyse networks of states and world regions :

But before to validate this results we need :

1. to clarify our definition of world regions and the associated list of target units.
2. to enlarge the dictionary to other languages (turkish, arabic, english)

Objectives

The objective of this short note is to explore the possibility of Wikidata for the production of multilingual dictionaries of world regions and more generally regional imaginations. In order to test the interest of this approach, we will try to produce multilingual dictionaries for the identification of different types of “regions” related to the division of the Earth (“natural”) or the division of the World (“political”)

• “Earth/Natural regions” : i.e. regions that are used in Atlas of Dictionaries for the localization at the surface on the earth and/or taught to young students in textbook as primary or secondary divisions of the surface of the Earth.

• “World/Political regions” : i.e. groups of states that are linked by formal international organization (e.g. European Union, ASEAN, …) and/or are perceived as linked politically or economically by themselves of in the eye of external observers (e.g. BRICS, Eurasia)

Earth/Natural regions: Atlas

So-called “Physical maps” in Atlas are a good source :

• continents : Asia, Africa, Europa, …
• subcontinents : North Africa, Western Europe,..
• peninsula/ archipel/ … : Scandinavia, Caraïbes, …
• bioclimatic area : Sahara, Amazonia, Sahel
• mountains/plain/… : Alps, Balkans, Ands, …
• Oceans/Seas/Gulf : Atlantic, Arctic, Mediterranea

Earth/natural regions : Textbooks

Textbooks and educative games for children are also crucial :

World/Political regions : IGO

An attempt to classify intergovernmental organization in 4 types :

1. Continental organizations
2. Subcontinental organizations
3. Transcontinental alliances
4. Heritage of empires

Source : https://commons.wikimedia.org/wiki/Atlas_of_international_organizations

World/Political regions : Other …

A cross-language perspective

We propose to etablish a dictionary of Earth and World Regions in the five languages of interest for the project IMAGEUN :

• english : applied to media of UK and Ireland
• french : applied to media of France and Tunisia
• german : applied to media of Germany
• turkish : applied to media of Turkey
• arabic : applied to media of Tunisia

We want to avoid any “eurocentric” or “anglocentric” perspective in the definition of entities. Therefore our definition of entities will follow the following rules :

1. Non universal : Entities will not necessary be available in all languages
2. Non equivalent : Translation of names does not imply equivalence of entities
3. Non hierarchic : An entity has different definitions in each language. None of the language can be considered as “pivot” or “reference.”

Entities equivalences and lexical universes

To summarize, we propose to build partial equivalences between entities that belong to different lexical universes.

The comparison between lexical universes will be necessarily limited to a small sample of entities for which we can assume that the entities are approximately equivalent.

2. WIKIDATA

Wikidata defines itself as

• a free and open knowledge base that can be read and edited by both humans and machines.
• as central storage for the structured data of its Wikimedia sister projects including Wikipedia, Wikivoyage, Wiktionary, Wikisource, and others.
• a support to many other sites and services beyond just Wikimedia projects.
• The content of Wikidata is available under a free license, exported using standard formats, and can be interlinked to other open data sets on the linked data web.

Codification of entities

The first interest of wikidata is to provide unique code of identifications of objects. For example a research about “Africa” will produce a list of different objects characterized by a unique code :

Informations on entities

Once we have selected an entity (e.g. Q15) we obtain a new page with more detailed informations in english but also in all other languages available in Wikipedia.

Informations on entities

A lot of information are available concerning the entity but, at this stage, the most important ones for our research are :

1. the translation in different languages
2. the equivalent words or expression in different languages
3. the definitions in different languages
4. the ambiguity of the term in each language and the potential risks of confusion with other entities.

Of course we should not take for granted the answers proposed by wikidata (as noticed by Georg, Wikipedia is a matter of research for IMAGEUN …) but without any doubt, it offers a very good opportunity to clarify our questions and help us to build tools for recognition of world regions and other geographical imaginations in a multilingual perspective.

Multilanguage defintions

A wikipedia entity like Q15 is an element of an ontology designed by its author for specific purposes. The specificity of the wikidata ontology is the fact that it is a multilinligual web where Q15 is a node of the web present in different linguistic layers. It means that we don’t have a single name or a single definition of Q15, except if we adopt the neocolonial perspective to choose the english language as reference. Depending on the context (i.e. the language or sub-language), Q15 could be defined as :

Correspondance between entities ?

The existence of the same code of wikipedia entities does not offer any guarantee of concordance between the geographical objects found in news published in different languages or different countries. But - and it is the important point - it help us to point similarities and differences between set of geographical entities that are more or less comparable in each language.

• Ex. 1 Amazonie : In french language, Amazonie is associated to the entity Q2841453 which is defined as a “région naturelle en Amérique du Sud”. But this entity does not exist apparently in turkish language. At the same time the french language propose also an entity “Forêt amazonienne” Q177567 defined as forêt équatoriale située dans le bassin amazonien en Amérique du Sud which is present in tukish language. We have also a third entity bassin amazonien identified as Q244451 which refers to so-called régions naturelles based on water basin (Cholley 1939)

• Ex.2 Proche-Orient/Moyen-Orient/Pays du Golfe/ Asie de l’Ouest : In french language we find four entities describing the complex geopolitical area located in western part of Asia and eastern part of Mediterranea. But this entities are not necessary used in all languages with the same frequency and can be completed by other entities like western Asia. The entity Proche-Orient (Q48214) which is frequent in french is only available in 30 languages when the Moyen-Orient (Q7204) is available in 84 languages, golfe Persique (Q34675) in 77 languages and Asie de l’Ouest (Q27293) in 71 languages.

Cross-language perspective

Having in mind the limits of the equivalence of entities across languages, it can nevertheless be an interesting experience to select a set of wikipedia entities (Q15, Q258, Q4412 …) and to examine their relative frequency in our different media from different countries with different languages. A typical hypothesis could be something like :

• Is Q15 more mentionned than Q46 in Tunisian newspapers ?

which is not equivalent to the question

• Is Africa more mentionned than Europe in Tunisian newspapers

but rather equivalent to the two joint questions

• Is the textual unit “Afrique” more mentionned than the textual unit “Europe” in Tunisian newspapers published in french language.
• Is the textual unit “إفريقيا” more mentionned than the textual unit “أوروبا” in Tunisian newspapers published in arabic language.

Boring details

The package WikidataR is an interface for the use of the Wikidata API in R language. Equivalent tools are available in Python and other languages for those non familiar with R. And it is of course possible to use directly the API. The first step is to install the most recent version of the R package WikidataR which install also related packages of interest.

#install.packages("WikidataR")
library(WikidataR)

Boring details

If we start our research with the word “Afrique” in french language we find more than 50 entities that contain this word in their label. Only the first 10 are presented below :

Boring details

The analysis of the list of result reveals four situations :

1. Target entities: A first list is related to entities that can be considered as world regions or geographical imaginations of interest for IMAGEUN. It is typically the case for the whole continent of Afrique (Q15) and its different subdivisions like North Africa (Q27381), West Africa (Q4412), Sub-Saharan Africa (Q132959).

2. Control entities : A list of entities that are not regions but should be controled if we want to identify our target entities. A typical example is the sovereign state of South Africa (Q258) which will necessary introduce mistakes in the identification of Africa as a continent if it is not controled. The problem will not necessary exist in all languages (e.g. German) but is important.

3. Ambiguous entities : Some entities are ambiguous because they are not regions but use exactly the same textual units than a target entity. It is for example the case of the roman province of Africa (Q181238) which can not be easily differentiated from the continent, except by manual inspection. This units are not easy to control but fortunately are generally not frequent.

4. Insignificant entities : Those entities that are exceptional inthe corpus can be simply gnored.

Workflow in a nutschell

We propose a semi-automatic method of extractions of entities in different languages that implies the presence of human expert at each step of the analysis. The figure below describe an example of research of world regions related to Africa in three languages.

The programs used for computer implementation are explained in the media cookbook on github with an example of implementation available onf the following page

3. Q & D EXPERIMENTS

We have realized a test of the previous workflow on an arbitraty selection of entities which are mainly related to continent and “natural” Earth divisions :

• continents : Asia, Africa, Europa, …
• subcontinents : North Africa, Western Europe,..
• peninsula/ archipel/ … : Scandinavia, Caraïbes, …
• bioclimatic area : Sahara, Amazonia, Sahel
• mountains/plain/… : Alps, Balkans, Ands, …
• Oceans/Seas/Gulf : Atlantic, Arctic, Mediterranea, …

This quick and dirty analysis does not offer any guarantee of quality because :

1. The list of entities has not been discussed by the IMAGEUN’s partners
2. The dictionary established in the different languages has not been controled by native speakers

The purpose is therefore only to provide food for thought.

Data

We start from a corpus of text where target wikipedia entities has been recognized :

Experience 1 : An Inter-Language analysis of lexical universes

Experience 1 : Europe (Q46) in 2019

Experience 1 : Europe (Q46) in 2020

Experience 1 : Mediterranea (Q4918) in 2019-2021

Experience 2 : A Cross-Language analysis of regional entities

Experience 2 : Data aggregation

For the experience 2, we create a new object called hypercube where the text of news has been removed and where we keep only the number of tags or proportion of news speaking from one or several regions (where1, where2), by media (who) and by time period (when)

## Joining, by = "id"

Experience 2 : Top 20 regions in full corpus

We can propose firstly a table of top entities in the whole corpus of newspapers.

• NB : all the names are present in the different languages in order to avoid eurocentrism or anglocentrism.

Experience 2 : Turkish newspapers - Top 10 regions

Experience 2 : German newspapers - Top 10 regions

Experience 2 : French newspapers - Top 10 regions

Experience 2 : UK newspapers - Top 10 regions

Experience 2 : Irish newspapers - Top 10 regions

Experience 2 : Tunisian Newspapers- Top 5 regions

Due to the limited number of news, only top 5 news is presented. The newspaper Babnet was in arabic language.

Experience 2 : Correspondance analysis - Factor 1-2

## Joining, by = "id"

Experience 2 : Factors 3-4

Experience 2 : Cluster analysis(world regions)

Experience 2 : Cluster analysis (medias)